Measurement instruments and assessment of study variables

Assessment of heart rate variability

In this study, HRV was assessed over 24 hours, while participants performed their daily activities. Compared to short-term HRV recordings, the long-term HRV assessment is commonly known as the gold-standard. However there are plenty of studies, where HRV is assessed using short-term ECG recordings of 2, 5, 10 minutes or even 2 hours. A common convention about the duration of short-term recordings does not exist. However, at least 2 minutes of recording are necessary to assess the HF and LF components of HRV.43 Since total variance of HRV is changing with the duration of ECG recording, the absence of a standardized short-term recording duration might be a disadvantage. To compare absolute HRV values across different individuals is therefore difficult.43 Based on the change in total variance, the distribution of frequency-domain variables is markedly different in short-term recordings compared to long-term HRV recordings. In any case, one should be careful when comparing results of HRV analysis with different ECG recording durations and should be aware of this problem. However, independent of the recording length, short-and long-term HRV has shown to be associated with cardiovascular risk and events.47,48,58

Another important point to consider is to use of a well-functioning software to edit ECG recordings. Artefacts may have a crucial influence on HRV values, which makes the editing process enormously important. This process includes the removal of artefacts but also the redefinition of premature ventricular and atrial beats. The quality of the Holter ECG recordings in our study was very high, which resulted in an editing time of 10.2 minutes per Holter-ECG. Two researchers were trained by a cardiologist to perform the editing process. Trainings were regularly repeated in order to maintain high quality standards. Difficult cases were discussed together with an experienced electrophysiologist.

8.DISCUSSION

102

Due to feasibility reasons, study participants carried the 24-h ECG device simultaneously with a 24-h BP monitoring device. To wear both devices simultaneously could be impairing for some participants. Therefore, it is possible that the behavior of some individuals was different compared to normal days. Unfortunately, there are no possibilities of controlling for this behavior. Nevertheless, we are confident that this phenomenon is attenuated with increasing sample size.

The task force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology recommend a recording duration of at least 18 hours for long- term ECG recordings.43 In our study we have defined a cut-off of at least 80% of the maximal duration of 24 hours, which results in a cut-off of 19.2 hours. Participants with recording durations below this cut-off were asked and motivated to repeat the recording. Even though overall quality of the ECG recordings was very high, we had to exclude ECG recordings of 22 participants. Moreover, 20 participants have denied to perform a 24-h ECG or had technical device issue during 24-h ECG recording.

Assessment of physical activity and other lifestyle factors

Lifestyle factors, such as physical activity, diet and smoking status were self-assessed using standardized questionnaires. Questionnaires are commonly used in epidemiological studies with a large sample size and are a helpful tool to assess different items. Using questionnaires has several advantages and disadvantages, which should be taken into account before deciding for or against their use. Important advantages are the low costs, the easy way of distribution as well as the potential of saving time. Moreover, the questionnaires used in the GAPP-study were scanned and data were automatically entered in the dataset. One main disadvantage of using questionnaires is that the assessment is subjective and there remain a potential uncertainty. Recall and response bias, which also includes social desirability bias, may occur and are difficult to control.111,112 Social desirability bias is based on the phenomenon that some individuals tend to answer what they think is more esteemed in the society. As an example, individuals might overestimate their physical activity because they know that being physically active is healthy. Even though these sources of uncertainty when using questionnaires, we expected that these potential biases can be minimized when increasing the number of study participants.

Diet was assessed using the official questionnaire of the Federal Office of Public Health (Swiss health survey 2007), where information about the frequency of fruit, vegetable, meat, fish or dairy consumption was assessed. While diet and smoking status are often ascertain

8.DISCUSSION

103

using questionnaires there are several options to assess physical activity.113 The choice of one specific method to assess physical activity is depending on the research question and the resources and possibilities of the study. The short IPAQ is an international and commonly used questionnaire assessing habitual moderate and vigorous physical activity and time spent for walking and sedentary activities. The validity and reliability of the short IPAQ was evaluated in an international population aged 18 to 65 years and achieved acceptable measurement properties.83 As always with questionnaires, one should be aware that collected data remain an approximation of the true result. The IPAQ is assessing information about the frequency and duration of physical activity, which could be misjudged by study participants. Moreover, seasonal changes might play an important role and could be included in the estimation of the participants.114 Underreporting may occur if individuals do not remark physical activity (mainly moderate physical activity) in their daily routine. In contrast, individuals may include socializing, theoretical instructions or refreshments in the physical activity duration.

Lifestyle and cardiovascular health metrics score

Adopting a healthy lifestyle is highly promoted by professional societies and governments.6,7 To investigate associations of a healthy lifestyle as a whole construct with different outcomes is therefore essential. As a consequence, different scores, based on lifestyle factors and/or cardiovascular health metrics, were constructed. Up to now, there is no consensus for the use of one specific score.

For our project, we have chosen a score, which is based on the American Heart Association (AHA) definition of ideal cardiovascular health metrics. As reported in the manuscript in chapter 4, this score consists of a broad set of cardiovascular health metrics and lifestyle factors, such as smoking, physical activity, diet, BMI, blood pressure, total cholesterol and HbA1c levels. Until now, this score was used in different US cohort studies with middle-aged

adults from the general population, where they showed a strong association with cardiovascular outcomes and mortality.8,9 However, there remain several issues concerning the score that should be discussed in the following section.

BMI is a widely used variable to assess body weight in relation to height with a universal classification independent of sex. The most important disadvantage of the BMI is that body composition is not taken into account. Based on the higher average muscle mass of men compared to women, BMI in men is often overestimated. Using the BMI as a risk predictor is therefore not always reliable. Sex-specific calculations or cut-offs might be one solution.

8.DISCUSSION

104

WHR for example, an easily measurable marker, takes into account the distribution of body fat. Universal and sex-specific cut-offs for WHR exists. Several studies showed that WHR is a stronger predictor for future cardiovascular events compared to BMI.115-117 Therefore, including WHR instead of BMI into the score should be considered as a meaningful alternative.

Total cholesterol is one of seven components of the lifestyle-score. Taking into account the strong association of LDL with cardiovascular outcomes,118 the inclusion of LDL instead of total cholesterol could be a good option. For a sensitivity analysis, we have modified this score accordingly, using a LDL cut-off of 160 mg/dl. As presented in chapter 4, this modification had no consequences on the relationship between HRV and the lifestyle-score. High BP is one of the most important cardiovascular risk factors. To include this strong risk factor in the score is therefore justified. However, it could be discussed whether conventional BP levels, measured in the doctor’s office, or ambulatory 24-h BP levels are more useful and reliable. Ambulatory 24-h BP levels have shown to be a stronger predictor for cardiovascular outcomes compared to conventional BP.119,120 Another study showed a high proportion of young individuals with a masked hypertension (27% among men), meaning that they have normal conventional BP, but an elevated ambulatory BP.121 With regard of our young study population, a remarkable proportion of our study participants could be misclassified. Nevertheless, compared to conventional BP measurement, ambulatory 24-h BP monitoring needs markedly more time and financial resources and it seems therefore not realistic to commonly include 24-h ambulatory BP levels into the score.

In the original AHA based score, diet relies on five dietary items, including salt consumption, fish intake, fruit- and vegetable intake, consumption of sweet beverages and whole grain consumption. The questionnaire we used for our study does not assess information about the consumption of sweet beverages and whole grains. This forced us to modify the definition of an ideal diet. We are aware of this problem and cannot rule out that this modification might have an influence on our study results.

Assessment of sleep-related breathing disorders

Polysomnography is the current standard diagnostic tool for the diagnosis of sleep-related breathing disorders. This procedure is rather complex and expensive and is done stationary in a sleep laboratory. Polysomnography is a multichannel recording of electromyographic, electroencephalographic and respiratory activity together with an ECG to detect any breathing disorders during sleep.16 Because of the financial resources, the sample size and

8.DISCUSSION

105

feasibility of the study, we have decided to instead perform a nighttime pulse oximetry including nasal airflow measurement, which can be handled by the study participants independently.

This portable device (ApneaLink, Resmed, USA) was validated in a validation study and data were compared to simultaneously recorded polysomnographic data.122 AHI and ODI assessed using the ApneaLink device were compared to the AHI based on the polysomnography. The sensitivity for AHI was ≥ 90% with a very high specificity. The ODI had a good overall sensitivity (≥80%). However, there was a lower specificity, especially at lower polysomnographic AHI levels. This may lead to a higher number of false positive results.122 In our analysis (chapter 5), 120 individuals had an AHI≥5 while 141 participants had an ODI ≥5. This difference might partly be reasoned by the lower specificity of the ODI. Despite we have not used the gold-standard method for the assessment of sleep-related breathing disorders, we are confident that the usage of the ApneaLink device including pulse oximetry and nasal airflow was the best possible solution to investigate our study research questions.

Nocturnal pulse oximetry and nasal air flow measurement was routinely implemented about one year after the official start of the baseline examination. Therefore, 610 participants were included in the study before having initiated this measurement. In addition, 127 participants have declined to perform this analysis or had artificial fingernails, which did not allow the nocturnal pulse oximetry. Unfortunately, among several participants the recording duration of either nocturnal pulse oximetry or nasal airflow measurement was too low, leading to the exclusion of these recordings (n=114). One reason for a shorter recording duration was a shift of the nasal cannula in order that nasal airflow could not be measured. To avoid this problem, we have instructed participants to fix the nasal cannula with tape on both cheeks. A few participants reported that they could only breathe through the mouth, which of course resulted in no recording time of the air flow measurement. Other participants reported of having lost the pulse oximetry device during night. Overall, the device was well tolerated of the participants and easy to handle.

8.DISCUSSION

106